1. Introduction and problem

In the 3-to-6-year span —kindergarten, preschool, CENDI, infant school— a package of artifacts has been installed that claim to count as support for guided play: a chatbot or a GenAI that produces guided play scripts, guided play lesson packs, play provocation banks or adult move checklists by prompt as a closed recipe; a classifier, an ML system or a camera that assigns a guided play quality score, adult guidance fidelity, play guidance compliance or minutes of guided play/hour to the child or the teacher; an administrative dashboard of guided minutes or provocation completion rate; a robot or an agent presented as a substitute that «guides play» without contingent adult mediation; and a ranking of teacher scaffolding effectiveness derived from computer vision or NLP without situated pedagogical judgment. The artifacts make it possible to display that the center «already does guided play with AI». The leap —from script, score, count, substitute-robot or ranking to claiming support— is not authorized by the mappings of AI in early childhood education nor by pedagogy when there is an adult who contingently guides the child's play, with curricular goals, child agency and situational scaffolding, on the free–guided–directed continuum, with co-presence and pedagogical interpretation, not a «guided play quality score».

The thesis is restrictive: those artifacts do not constitute support for guided play in early childhood. In early childhood education that practice develops when the adult has a learning goal, the child retains choice and agency, and guidance is flexible —open questions, hints, modeling, co-play, adjusted scaffolding— so that the adult notices, interprets and responds (Skene et al., 2022; Wickstrom y Pyle, 2024; Sanchez, 2025; Cekaite y Simonsson, 2025); not a GenAI of scripts, a quality score, a dashboard of minutes, a robot that «guides play» nor a scaffolding ranking. AI may support the adult's preparation, documentation and reflection in a subordinate way (Jambunathan, 2025; Omoeva et al., 2024; Li, Fleer y Yang, 2021; Lee, Lee y Yoon, 2025); it must not substitute contingent co-construction nor convert guidance into a surveillance metric. Skene et al. (2022) synthesize the experimental evidence and underline the heterogeneity of conceptualization and implementation: situated craft, not an auto-score. Wickstrom y Pyle (2024) describe a continuum —extend, facilitate, invite— before, during and after play. Sanchez (2025) documents a teacher inquiry on scaffolding writing in guided play. That authorizes asking what was measured: a quality score or a fidelity ranking —not the practice when an adult holds the goal, protects agency and scaffolds, and a professional interprets the episode.

The problem is aggravated by ten category confusions. First: it is not free play —25 August covered that axis; here contingent guidance on the continuum, not the absence of an adult—. Second: it is not outdoor, loose parts, sand, water, playdough nor risky play. Third: it is not symbolic play nor spatial play as axis. Fourth: it is not co-regulation (11 September). Fifth: it is not SST (10 September). Sixth: it is not ToM (9 September), prosocial (8 September) nor peer-conflict (7 September). Seventh: it is not mark-making, dialogic reading nor literacy as axis —Sanchez (2025), Masters et al. (2023) and Cun (2025) are read for contingent guidance—. Eighth: it is not numeracy nor STEM as axis —Wickstrom y Pyle (2024) are read for the continuum; Critten et al. (2021) for adult guidance, with transfer—. Ninth: it is not inclusion as axis —Danniels y Pyle (2022, 2023) and Cade (2023) are read for the adult's role in play—. Tenth: it is not CLASS, executive functions nor formative assessment —Omoeva et al. (2024) are read as a boundary: formative toolkit ≠ surveillance score—. The artifacts fit on a slide; the episode in which an adult observes, asks and yields, does not. Contributions: separating craft and artifacts; examining marked cases; and offering four tests of support.

2. State of the art: situated practice vs artifact

It is useful to separate four strata that the market of «AI for guided play in early childhood education» usually mixes. The first is the construct from 3 to 6 years as situated relational practice: a free–guided–directed continuum in which the adult holds a learning goal, the child retains agency and guidance is flexible and contingent —the adult notices, interprets and responds to the child's cues— (Skene et al., 2022; Wickstrom y Pyle, 2024; Sanchez, 2025; Cekaite y Simonsson, 2025). The second is the craft that cultivates it —an adult who does not close the episode with a recipe; varies extend, facilitate and invite; enters and leaves the play frame; co-presence that interprets the episode, not only «there were guided minutes»— (Danniels y Pyle, 2022, 2023; Pyle et al., 2022; Li, Fleer y Yang, 2021; Masters et al., 2023). The third is the evidence of AI affordances in ECE, robots, GenAI, AI literacy and postdigital critique —without equating them to situated guided-play pedagogy— (Su y Yang, 2022; Su y Zhong, 2022; Lee et al., 2024; Lee, Lee y Yoon, 2025; Edwards, 2023; Jambunathan, 2025; Ormanci, 2026; Giray, 2026). The fourth is the framework of human measure and reflection, which treats the young child as a subject who plays with another, not as a vector of minutes of guided play/hour nor as the addressee of a robot that «guides play» (Omoeva et al., 2024; Pyle et al., 2022; Giray, 2026). Inference: a guided play quality score does not formulate the question that sustains the flow; an adult who guides contingently does. Su y Yang (2022) map AI without equating it to situated guided play. Lee et al. (2024) propose incorporating robots, not substituting mediation. Giray (2026) establishes that the authentic relationship is not replicated by a chatbot. Lee, Lee y Yoon (2025) find that teachers emphasize human interactions.

The verified systems measure objects distinct from the craft: effects of guidance during play (Skene et al., 2022), a continuum of approaches (Wickstrom y Pyle, 2024), teacher inquiry (Sanchez, 2025), interactional strategies (Cekaite y Simonsson, 2025), play-based learning profiles (Pyle et al., 2022), the adult's role in inclusive play (Danniels y Pyle, 2022, 2023), playful vocabulary experiences (Masters et al., 2023), heritage language (Cun, 2025), PD to bring play into the program (Li, Fleer y Yang, 2021) and a formative toolkit (Omoeva et al., 2024). Inference: an adult guidance fidelity does not scaffold the move that the episode requires; an adult who interprets the situation does. Critten et al. (2021) and Cade (2023) saturate transfer anchors. Edwards (2023) anchors the postdigital framework: AI is not a neutral add-on to the kindergarten.

3. Review method

A critical narrative review was conducted, not a primary meta-analysis. The purpose was not to estimate a homogeneous effect size of the artifacts, but to articulate an argument of pedagogical category with verified sources. Inclusion criteria: (a) 2021–2026, with marked transfer when the sample does not equal 3–6 in a guided-play classroom, is primary, HRI/robot as framework, AI curriculum, community heritage language, special needs as an inclusion axis, programming/STEM, or a formative tool without equating to automatic pedagogy; (b) guided play, free–guided–directed continuum, adult scaffolding in play, play-based learning profiles, or AI in ECE / robots / GenAI / postdigital critique with artifact/craft relevance; (c) kindergarten, preschool, CENDI or 3–6, or explicit transfer; (d) peer-reviewed journal, chapter or report with DOI; (e) verifiable DOI or publisher page. Excluded as central object were axes already used in this series —free play, outdoor/loose parts, symbolic, spatial, co-regulation, SST, ToM, prosocial, peer-conflict, mark-making, dialogic reading, literacy, numeracy/STEM as axis, FE, CLASS, inclusion, formative assessment and documentation as axes—. Sanchez (2025) is read for scaffolding in guided play; Masters et al. (2023) for playful experiences; Wickstrom y Pyle (2024) for the continuum; Critten et al. (2021) for adult guidance (STEM and age transfer); Danniels y Pyle (2022, 2023) and Cade (2023) for the adult's craft in play. Omoeva et al. (2024) = formative toolkit ≠ surveillance score. Skene et al. (2022) = 1–8 years: age transfer.

The search was executed on 14 September 2026 (slot 09:02 America/Mexico_City, weekday routine) on DOI pages, Crossref, Springer, Elsevier, Wiley, Taylor & Francis, Frontiers, SAGE and publisher sites. Each source was verified against fuentes.md and crossref_check.json. Empirical finding, framework and pedagogical inference were distinguished. No N, d, r, AUC or DOI was invented: when an artifact lacks a verified trial as a display package in CENDI 3–6, it is discussed as a category ceiling. Twenty-one sources from the verified corpus were used.

4. Case 1. The artifacts do not constitute support

Su y Yang (2022) and Su y Zhong (2022) saturate the ceiling of artifacts when AI in ECE is presented as if it were support for guided play. Su y Yang (2022) —a scoping of seventeen studies, 1995–2021— map tools and impacts of AI for children (AI concepts, robotics, creativity, literacy, computational thinking), not that a guided play quality score cultivates contingent guidance. Su y Zhong (2022) propose an AI curriculum for kindergarten and a social robot as a learning companion of AI principles: a curricular framework, not guided play. Inference: «high quality score = there was support» is a ceiling of analytics. An adult guidance fidelity can coexist with an absence of child agency. The grammar of display inverts the sequence: first the classifier or the GenAI of scripts is activated; then it is declared that «there is already guided play with AI».

Lee et al. (2024) explore a practical framework for incorporating AI humanoid robots in early childhood education —appropriate concepts, child–AI relationship, activities and challenges—: a framework, not a trial that the robot «guides play» without adult mediation. Lee, Lee y Yoon (2025) —six early childhood teachers; 20–30-minute interviews— find that all six considered AI useful for learning and classroom management, but all emphasized human interactions and expressed concern about training and about AI making the teacher obsolete. Status: qualitative finding of perceptions. Inference: that voice prevents selling an agent that «guides play» without contingent mediation. Edwards (2023) situates AI in ECEC in a postdigital framework: it is not a neutral technical add-on. Ormanci (2026) —thirty-seven articles— finds an increase of interest after 2023, with ChatGPT as the most frequent tool, followed by robotic tools. Inference: that frequency does not authorize guided play scripts by prompt nor a substitute-robot. Giray (2026) argues that the role of AI in ECE for 0–7 years remains highly contested and that high-quality development requires authentic relationships that AI does not replicate: a critical framework. Restrictive inference: selling an agent that «guides play» in 3–6 confuses product with craft.

Jambunathan (2025) summarizes the use of GenAI (ChatGPT, Claude, Dall-E, among others) in early childhood teacher education for lesson plans: adult preparation. Marked inference: producing materials by prompt may be subordinate preparation; support begins when the adult notices, interprets and responds in the episode, not when a play provocation bank or an adult move checklist closes guidance as a recipe. The artifacts lack, in the verified corpus, pedagogical-equivalence trials with the craft in CENDI 3–6; no d, r or AUC of «support» is invented. They share a grammar of substitution: the GenAI closes; the score speaks for guidance; the dashboard substitutes observation; the robot erases mediation; the ranking converts process into a percentile without judgment.

5. Case 2. The relational craft in the kindergarten

The floor of the craft is not a score: it is an episode in which the adult holds a learning goal, the child retains choice and agency, and guidance is flexible —open questions, hints, modeling, co-play, adjusted scaffolding—; the adult notices, interprets and responds to the child's cues; co-presence that interprets play —not only «there were guided minutes»—; and, when the flow breaks, the adult yields or reformulates instead of imposing a checklist (Skene et al., 2022; Wickstrom y Pyle, 2024; Sanchez, 2025; Cekaite y Simonsson, 2025). Inference: there is support when that practice is protected; not when a quality score, a GenAI of scripts or a fidelity ranking is displayed. The craft is recognized in the episode: the child initiates play; the adult observes, formulates a question or offers a material that extends toward a goal without hijacking agency; the robot or the GenAI, if they appear, are subordinate means to the adult, not substitutes.

Skene et al. (2022) saturate the experimental-evidence anchor and must be read with age transfer. Systematic review and meta-analysis: 39 studies (1977–2020); 17 in the meta-analysis (N total = 3893; mean age from 1 to 8 years). Empirical finding: guided play had a larger positive effect than direct instruction on early mathematical skills (g = 0,24), shape knowledge (g = 0,63) and task switching (g = 0,40), and larger than free play on spatial vocabulary (g = 0,93); no differences were identified for other key outcomes. The narrative synthesis underlined heterogeneity in conceptualization and implementation. Three traits: the adult's learning goal; the child's agency and choice; flexible guidance techniques. Transfer: the 1–8 range does not equal CENDI 3–6; the meta-analysis does not authorize a quality score nor a dashboard of minutes. Marked inference: heterogeneity of implementation ≠ closed recipe; a lesson pack by prompt misreads precisely what the synthesis flags as variable.

Wickstrom y Pyle (2024) saturate the classroom-continuum anchor. Qualitative study of how kindergarten educators implement guided play to support early mathematical learning. Empirical finding: three approaches —extend, facilitate, invite— as adult facilitation and child autonomy vary; guided play is situated in an iterative process in which the educator intervenes before, during and after play to consolidate, connect and deepen. Marked distinction: it is cited for the continuum of guidance, not as a numeracy axis. Inference: extend/facilitate/invite is situated craft, not adult guidance fidelity; «before, during and after» cannot be reduced to minutes of guided play/hour. Sanchez (2025) saturates the teacher-inquiry anchor: a five-month participatory action research in a public kindergarten classroom; methods for incorporating writing instruction through guided play in four classroom spaces, with materials and scaffolding strategies. Status: finding of practice and inquiry. Marked distinction: it is read for contingent scaffolding of writing in play, not as mark-making nor dialogic reading. Inference: curating materials and dialoguing in the episode ≠ a GenAI of guided play lesson packs.

Cekaite y Simonsson (2025) saturate the interactional anchor. Empirical chapter on educators' strategies in adult–child play in Swedish preschool; they analyze teacher questions as a resource to initiate, sustain and guide play. Finding: some question formats sustain flow and agency; others displace the focus or close play; «out-of-frame» talk can open a space of social and linguistic learning; they warn of stumbles in egalitarian dialogue and underline a role of democratic guidance. Status: finding of interactional analysis. Restrictive inference: an adult move checklist by prompt does not distinguish the question that sustains the flow from the one that breaks it; that requires co-presence. Danniels y Pyle (2022) —multiple-case study; three «enactor» kindergarten teachers who valued inclusion in play; observation (a minimum of three hours per classroom) and interviews in autumn and spring— find a balance between child agency and teacher guidance, child-centered and flexible involvement, and support for social interactions; observed practices: one-to-one conversations, small groups, co-play and collaborative problem-solving. Marked distinction: it is read for the craft of guiding play, not as an inclusion article. Danniels y Pyle (2023) —eight kindergarten teachers in Ontario; children with neurodevelopmental delay or disability— explore inclusion in play: indirect environmental strategies and proactive teacher support for social participation. Inference: the craft is contingent and situated; not a play guidance compliance.

Pyle, DeLuca, Wickstrom y Danniels (2022) —a two-phase qualitative approach; observation and interviews in 18 kindergarten classrooms— find three play profiles aligned with assessment approaches: free play–informal assessment; teacher-directed play–formal assessment; play continuum–blended assessment. Contrast inference: human profiles of practice ≠ a scaffolding ranking by vision/NLP; blended assessment is professional judgment, not a quality score. Masters et al. (2023) propose a toolbox of playful vocabulary experiences —active, engaging, meaningful, socially interactive, iterative and joyful— as a complement, not a substitute, of shared reading. Marked distinction: it is read for the design of experiences with an adult, not as literacy. Inference: a toolbox for the craft ≠ a play provocation bank closed by prompt. Cun (2025) —qualitative study in a community classroom of Chinese heritage language; observations, video and one interview with the teacher— finds that, with a scarcity of physical materials, the teacher created slides as digital resources for play scenarios in which children used vocabulary and phrases. Transfer: community heritage classroom, not CENDI 3–6. Inference: a digital resource subordinate to the teacher ≠ a GenAI of scripts.

Li, Fleer y Yang (2021) —video of a kindergarten teacher in China and interviews; Educational Experiment with Conceptual PlayWorld— argue that the educational experiment creates a new social situation, amplifies reflective practice and operates as a source of PD to bring play practices into the program. Inference: educating the adult ≠ an adult move checklist; collective reflection is not a dashboard of provocation completion rate. Cade (2023) —exploratory case; seven participants (six preschool teachers and one coach) in New Jersey; interviews and lesson plans, without classroom observation— finds that some teachers believe children with special needs do not belong in the general classroom; they describe teacher-directed grouping and a lack of training; a finding of concern is the description of practices such as restraining young children. Distinction: it is read as a contrast of absent craft, not as an inclusion article nor as a trial of successful guided play. Inference: a GenAI of lesson packs does not repair the absence of craft. Critten, Hagon y Messer (2021) —fifteen children aged 2 to 4 years; three coding clubs; six sessions of 45–60 minutes; Bee-Bots— find incipient development of computational thinking, but the children needed adult support to direct the robots. Transfer: includes 2 years; STEM axis. Marked inference: the robot did not «guide play» —the adult did—. Taken together: none of these systems signs a GenAI of scripts, a quality score, a dashboard of minutes, a substitute-robot or a teacher ranking. The evidence is real in its domain; «score / script / robot / ranking = support for guided play» is an illegitimate category inference.

6. Case 3. Category boundaries

Ten boundaries protect the axis. First: free play —Skene et al. (2022) contrast guided play with free play; Wickstrom y Pyle (2024) situate extend/facilitate/invite on a continuum; 25 August covered free play—. Inference: «we let them play = there is guided play» does not sign. Second: outdoor / loose parts / sand / water / playdough / risky play. Third: symbolic / spatial play —Skene et al. (2022) report spatial vocabulary as an outcome; this article is not converted into a spatial axis—. Fourth: co-regulation (11 September). Fifth: SST (10 September) —Cekaite y Simonsson (2025) analyze questions in play; the axis is guidance, not sustained shared thinking—. Sixth: ToM, prosocial, peer-conflict. Seventh: mark-making / dialogic reading / literacy —Sanchez (2025), Masters et al. (2023) and Cun (2025) are read for contingent guidance—. Eighth: numeracy / STEM —Wickstrom y Pyle (2024) are read for the continuum; Critten et al. (2021) for adult mediation, with transfer—. Ninth: inclusion —Danniels y Pyle (2022, 2023) and Cade (2023) are read for the adult's craft in play—. Tenth: CLASS, executive functions, formative assessment and documentation —Pyle et al. (2022) anchor human profiles; Omoeva et al. (2024) anchor the decisive boundary: Teacher RePlay is a formative toolkit of observation and reflection (protocol, coaching tips, Children ReAct), piloted in Bangladesh, Colombia and Uganda, for 3–12 years (stage and context transfer); it is not a guided play quality score nor a dashboard of guided minutes. Converting that formative tool into compliance surveillance inverts its meaning. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: coherent AI in 3–6 remains on the adult's side (preparation, documentation, reflection), not as a GenAI of lesson packs, a quality score, a dashboard, a robot that «guides play» nor a ranking without judgment.

7. Four tests of support (not artifact)

The framework that follows is a pedagogical inference of this article, anchored in the cases and in the verified instruments. It is not a new international standard. It distinguishes four tests. If a kindergarten, preschool, CENDI or infant school does not pass them, it cannot declare that the artifacts constitute support for guided play in early childhood.

7.1. Test of situated practice (curricular goals, child agency and contingent scaffolding on the free–guided–directed continuum), not of the classifier/ML/camera nor of the robot/agent that «guides play». Skene et al. (2022), Wickstrom y Pyle (2024), Sanchez (2025) and Cekaite y Simonsson (2025) define the craft/artifact contrast. Su y Yang (2022), Lee et al. (2024) and Giray (2026) map AI and robots as affordance or framework, not as situated guided play. Critten et al. (2021, transfer) show that the Bee-Bot robot did not substitute adult guidance. Inference: if the «evidence» is a guided play quality score, adult guidance fidelity, play guidance compliance, minutes of guided play/hour or an agent that «guides play» without an adult, the center did analytics or HRI, not support.

7.2. Test of adult co-presence that notices, interprets and responds —not only «there were guided minutes»—, not of the GenAI of guided play scripts / lesson packs / play provocation banks / adult move checklists. Sanchez (2025), Cekaite y Simonsson (2025), Danniels y Pyle (2022, 2023) and Masters et al. (2023) situate practice and mediation. Jambunathan (2025) admits GenAI as adult preparation. Cun (2025, transfer) situates a digital resource subordinate to the teacher. Inference: a script by prompt does not sign the episode in which the adult distinguishes the question that sustains the flow from the one that closes it. Skene et al. (2022) show heterogeneity of implementation: a closed recipe misreads the craft.

7.3. Test of human mediation and of the interpretation of guided play as relational practice, not of the dashboard of guided minutes / provocation completion rate nor of the ranking of scaffolding effectiveness by vision/NLP. Pyle et al. (2022) anchor human profiles of practice and assessment. Omoeva et al. (2024) anchor the formative toolkit —observation, reflection, coaching tips, children's voices— with stage transfer (3–12) and context transfer; boundary: formative tool ≠ surveillance score. Li, Fleer y Yang (2021) anchor collective reflection as PD, not as a percentile. Inference: high guided minutes/hour do not by themselves raise scaffolding. Subordinate preparation of the adult is legitimate; the scorer that converts the child or the teacher into a vector of compliance is not.

7.4. Test of category distinction and of professional judgment, not of the product catalog. Guided play ≠ free play, outdoor/loose parts, symbolic/spatial, co-regulation, SST, ToM, prosocial, peer-conflict, mark-making, dialogic reading, numeracy/STEM as axis, inclusion as axis, CLASS nor automatic formative assessment. Sanchez (2025), Masters et al. (2023), Wickstrom y Pyle (2024), Danniels y Pyle (2022, 2023), Critten et al. (2021) and Omoeva et al. (2024) prevent those confusions. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: support is fulfilled with goals, agency and contingent scaffolding with co-presence that interprets —not by scoring guidance nor by substituting the adult with an agent. The framework admits digital subordinate to the adult's preparation, documentation and reflection (Jambunathan, 2025; Omoeva et al., 2024; Li, Fleer y Yang, 2021; Lee, Lee y Yoon, 2025). It rejects declaring support by the artifacts (Giray, 2026; Su y Yang, 2022; Lee et al., 2024; Ormanci, 2026). The four tests are read together.

8. Discussion

Three tensions. First: displaying the artifacts versus exercising support. The cases of sections 5–6 —with marked transfers and distinctions— sustain the contrast; Su y Yang (2022), Su y Zhong (2022) and Ormanci (2026) map AI, ChatGPT and robots without equivalence to situated guided play. Second: automated assessment or substitute-robot versus relational pedagogy —declaring support by a guided play quality score, adult guidance fidelity or by a system that «guides play» is inverted pedagogy; Pyle et al. (2022) show that even a human blended assessment is not the automatic craft; Omoeva et al. (2024) show that a formative toolkit is corrupted if it is converted into a compliance dashboard; Critten et al. (2021) show that the robot did not substitute the adult—. Third: a GenAI of provocation banks / a chatbot that «guides play» / a scaffolding ranking versus craft (Giray, 2026; Lee, Lee y Yoon, 2025; Jambunathan, 2025; Lee et al., 2024; Edwards, 2023): selling an agent as a guide of play in 3–6 confuses product with practice. Wickstrom y Pyle (2024) confirm that the continuum requires judgment; Cekaite y Simonsson (2025) confirm that the form of the question decides whether the flow is sustained; Sanchez (2025) confirms that scaffolding is designed in the classroom; Giray (2026) subordinates any use to the human relationship. Subordinate digital is legitimate toward the adult (preparation; Teacher RePlay as reflection; Jambunathan, 2025, as training; Li, Fleer y Yang, 2021, as PD); inverting the sequence is not. If the center shows goals, agency and contingent scaffolding, it may speak of support; if it only shows GenAI scripts, quality scores, guided minutes, robots that «guide play» or rankings without judgment, it speaks of artifacts.

9. Limits

This review is narrative. It does not apply its own PRISMA nor estimate combined effects. No d, r or AUC is invented beyond those reported by Skene et al. (2022) in their meta-analysis (1–8 years; 39 studies up to 2020: age and window transfer). Wickstrom y Pyle (2024) are qualitative on the continuum, not a numeracy axis. Sanchez (2025) is one teacher's inquiry in one classroom. Cekaite y Simonsson (2025) are an interactional chapter in Swedish preschool. Danniels y Pyle (2022, 2023) are three enactors and eight teachers in Ontario: they are not converted into an inclusion article. Pyle et al. (2022) measure 18 classrooms: profiles, not a quality score. Masters et al. (2023) are a vocabulary toolbox. Cun (2025) is a community heritage classroom: transfer. Li, Fleer y Yang (2021) analyze one teacher in China. Cade (2023) are seven participants without classroom observation. Critten et al. (2021) are fifteen children aged 2–4 years and Bee-Bots: STEM and age transfer. Omoeva et al. (2024) cover 3–12 years in three countries: formative toolkit, stage transfer. Lee, Lee y Yoon (2025) measure six teachers. Lee et al. (2024), Su y Yang (2022), Su y Zhong (2022), Edwards (2023), Jambunathan (2025) and Ormanci (2026) map AI/robots/curriculum/training/trends, not the display package in CENDI. Giray (2026) is a critical perspective. No verified trials of the artifacts as a single package of «guided play with AI» in a 3–6 classroom were located; they are discussed as a category ceiling. The inferences of section 7 are category hypotheses, not implementation evidence.

10. Conclusions

The artifacts —a GenAI of guided play scripts, lesson packs, play provocation banks or adult move checklists; a classifier/ML/camera of guided play quality score, adult guidance fidelity, play guidance compliance or minutes of guided play/hour; a dashboard of guided minutes or provocation completion rate; a robot/agent that «guides play» without contingent adult mediation; a scaffolding ranking by vision/NLP— do not constitute support for guided play in early childhood education. Su y Yang (2022) and Su y Zhong (2022) confirm AI affordances and curriculum without equivalence to situated guided play. Lee et al. (2024), Lee, Lee y Yoon (2025), Edwards (2023), Jambunathan (2025), Ormanci (2026) and Giray (2026) set frameworks, perceptions, postdigitality, teacher preparation, trends and critique, not a substitute-robot nor a GenAI of a closed recipe. The empirical cases (sections 5–6) require reading guidance during play, the continuum, teacher inquiry and the formative toolkit by what they measure and by what they do not equal as situated pedagogy. When there is support there are goals, agency and contingent scaffolding with co-presence that interprets (Skene et al., 2022; Wickstrom y Pyle, 2024; Sanchez, 2025; Cekaite y Simonsson, 2025). Guided play is distinguished from free play, outdoor/loose parts, symbolic/spatial, co-regulation, SST, ToM, prosocial, peer-conflict, mark-making, dialogic reading, numeracy/STEM as axis, inclusion as axis, CLASS and automatic formative assessment.

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure mappings of AI, robots, GenAI of training, a meta-analysis of guidance during play, a continuum, teacher inquiry, play-based learning profiles, PD or a formative toolkit, it does not translate them into support via quality score, script by prompt, dashboard of minutes, substitute-robot or teacher ranking. Accompanying girls and boys from three to six years in guided play is to exercise situated practices in which the adult contingently guides the child's play, with curricular goals, child agency and situational scaffolding, on the free–guided–directed continuum, and co-presence that interprets guidance as situated relational practice, not as a «guided play quality score». The rest is a GenAI of scripts, a quality score, a dashboard of minutes, a robot that «guides play» and a ranking. It is not support for guided play in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Cade, J. (2023). Child-centered pedagogy: Guided play-based learning for preschool children with special needs. Cogent Education, 10(2), Article 2276476. https://doi.org/10.1080/2331186x.2023.2276476
  2. Cekaite, A., y Simonsson, M. (2025). Guided play, child agency and learning: Educator strategies in adult-child play. En N. Pramling, A. Farrell, C. Wallerstedt y M. Nilsen (Eds.), Adults in play with children (pp. 101–119). Springer. https://doi.org/10.1007/978-3-031-99629-0_6
  3. Critten, V., Hagon, H., y Messer, D. (2021). Can pre-school children learn programming and coding through guided play activities? A case study in computational thinking. Early Childhood Education Journal, 50(6), 969–981. https://doi.org/10.1007/s10643-021-01236-8
  4. Cun, A. (2025). Chinese heritage language and literacy learning through guided play activities. Early Childhood Education Journal, 54(4), 3017–3025. https://doi.org/10.1007/s10643-025-02082-8
  5. Danniels, E., y Pyle, A. (2022). Inclusive play-based learning: Approaches from enacting kindergarten teachers. Early Childhood Education Journal, 51(7), 1169–1179. https://doi.org/10.1007/s10643-022-01369-4
  6. Danniels, E., y Pyle, A. (2023). Teacher perspectives and approaches toward promoting inclusion in play-based learning for children with developmental disabilities. Journal of Early Childhood Research, 21(3), 288–302. https://doi.org/10.1177/1476718x221149376
  7. Edwards, S. (2023). Early childhood education and care, artificial intelligence and the postdigital. Australasian Journal of Early Childhood, 48(3), 179–181. https://doi.org/10.1177/18369391231198910
  8. Giray, L. (2026). A critical stand against artificial intelligence in early childhood education. Discover Artificial Intelligence, 6(1), Article 768. https://doi.org/10.1007/s44163-026-01510-x
  9. Jambunathan, S. (2025). Integrating artificial intelligence into early childhood teacher education. Contemporary Issues in Early Childhood. https://doi.org/10.1177/14639491251340141
  10. Lee, J., Jo, J., Lee, J. O., y Kim, S. H. (2024). Incorporating humanoid artificial intelligence (AI) robots into early childhood education. Early Childhood Education Journal, 53(8), 2849–2857. https://doi.org/10.1007/s10643-024-01690-0
  11. Lee, J., Lee, J. O., y Yoon, J. (2025). Exploring perceptions of early childhood teachers on the use of artificial intelligence in early childhood education. Journal of Early Childhood Teacher Education, 1–16. https://doi.org/10.1080/10901027.2025.2600033
  12. Li, L., Fleer, M., y Yang, N. (2021). Studying teacher professional development: how a Chinese kindergarten teacher brings play practices into the program. Early Years, 42(1), 104–118. https://doi.org/10.1080/09575146.2021.2000942
  13. Masters, A. S., Scott, M. E., Wright, C. A., Spiewak Toub, T., Dickinson, D. K., Golinkoff, R. M., y Hirsh-Pasek, K. (2023). Playing with words: Using playful learning experiences in the early childhood classroom to build vocabulary. The Reading Teacher, 76(6), 775–783. https://doi.org/10.1002/trtr.2201
  14. Omoeva, C., Zosh, J. M., Pyle, A., D'Sa, N., Contreras Gomez, R., Dooley, B., Giacomazzi, M., Ariapa, M., Maldonado-Carreño, C., Escallón, E., Dey, G., Pavel, K. F., y Laverty, C. (2024). Teacher RePlay and Children ReAct: pilot testing a formative toolkit to support playful learning in the classroom. Frontiers in Education, 9, Article 1342424. https://doi.org/10.3389/feduc.2024.1342424
  15. Ormanci, U. (2026). Artificial intelligence in early childhood education: a systematic review of trends, methods, and recommendations for quality education. European Early Childhood Education Research Journal, 1–26. https://doi.org/10.1080/1350293x.2026.2651904
  16. Pyle, A., DeLuca, C., Wickstrom, H., y Danniels, E. (2022). Connecting kindergarten teachers’ play-based learning profiles and their classroom assessment practices. Teaching and Teacher Education, 119, Article 103855. https://doi.org/10.1016/j.tate.2022.103855
  17. Sanchez, A. (2025). Guided play in the kindergarten classroom: One teacher’s inquiry into scaffolding play-based writing instruction. Early Childhood Education Journal, 53(6), 2089–2098. https://doi.org/10.1007/s10643-025-01931-w
  18. Skene, K., O’Farrelly, C. M., Byrne, E. M., Kirby, N., Stevens, E. C., y Ramchandani, P. G. (2022). Can guidance during play enhance children’s learning and development in educational contexts? A systematic review and meta-analysis. Child Development, 93(4), 1162–1180. https://doi.org/10.1111/cdev.13730
  19. Su, J., y Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, Article 100049. https://doi.org/10.1016/j.caeai.2022.100049
  20. Su, J., y Zhong, Y. (2022). Artificial intelligence (AI) in early childhood education: Curriculum design and future directions. Computers and Education: Artificial Intelligence, 3, Article 100072. https://doi.org/10.1016/j.caeai.2022.100072
  21. Wickstrom, H., y Pyle, A. (2024). Supporting early math learning along a continuum of guided play. Early Childhood Education Journal, 53(4), 1053–1063. https://doi.org/10.1007/s10643-024-01650-8